Research Post
The power of associative classifiers is to determine patterns from the data and perform classification based on the features that are most indicative of prediction. Although they have emerged as competitive classification systems, associative classifiers suffer from limitations such as cumbersome thresholds requiring prior knowledge which varies with the dataset. Furthermore, ranking discovered rules during inference rely on arbitrary heuristics using functions such as sum, average, minimum, or maximum of confidence of the rules. Therefore, in this study, we propose a two-stage classification model that implements automatic learning to discover rules and to select rules. In the first stage of learning, statistically significant classification association rules are derived through association rule mining. Further, in the second stage of learning, we employ a machine learning-based algorithm which automatically learns the weights of the rules for classification during inference. We use the p-value obtained from Fisher’s exact test to determine the statistical significance of rules. The machine learning-based classifiers like Neural Network, SVM and rule-based classifiers like RIPPER help in classifying the rules automatically in the second stage of learning, instead of forcing the use of a specific heuristic for the same. The rules obtained from the first stage form meaningful features to be used in the second stage of learning. Our approach, BiLevCSS (Bi-Level Classification using Statistically Significant Rules) outperforms various state-of-the-art classifiers in terms of classification accuracy.
Feb 15th 2022
Research Post
Read this research paper, co-authored by Amii Fellow and Canada CIFAR AI Chair Osmar Zaiane: UCTransNet: Rethinking the Skip Connections in U-Net from a Channel-Wise Perspective with Transformer
Sep 27th 2021
Research Post
Sep 17th 2021
Research Post
Looking to build AI capacity? Need a speaker at your event?